JP5754504B2 - Management apparatus, information processing apparatus, information processing system, and data transfer method - Google Patents

Description

  The present invention relates to a management device, an information processing device, an information processing system, a data transfer method, a route selection program, and a transfer request program.

  Conventionally, a computer system having a plurality of nodes employs a “cluster configuration” in which nodes are connected by an interconnect. When the cluster configuration is adopted, the processing distributed at each node is executed in parallel, so that the performance of the computer system is improved. Further, even when a failure occurs in some nodes and the processing is stopped, the processing is continuously executed in the node where the failure does not occur, so that the availability of the computer system is improved.

  A configuration example of a computer system adopting a cluster configuration will be described with reference to FIG. FIG. 21 is a diagram illustrating a configuration example of a computer system adopting a cluster configuration. As illustrated in FIG. 21, the computer system 900 includes calculation nodes 901 to 912, IO (Input Output) nodes 913 to 916, and IO devices 917 to 918. The calculation nodes 901 to 912 execute various calculations and application processes. The IO nodes 913 to 916 control input / output to / from the IO device. The IO devices 917 to 918 are storage devices that store data, applications, and the like, for example.

  As shown in FIG. 21, each computation node and each computation node, each IO node and each IO node, and each computation node and each IO node are connected by an interconnect. In the example of FIG. 21, the IO node 913 is connected to the IO device 917, the IO node 914 is connected to the IO device 918, and the IO node 915 is connected to the IO device 918. The IO node 916 is connected to the IO device 917. Note that the connection between the IO node 914 and the IO device 918 indicated by a dotted line is a standby connection that is operated when an abnormality occurs in the connection between the IO node 915 and the IO device 918 indicated by a solid line. Similarly, the connection between the IO node 916 and the IO device 917 indicated by a dotted line is a standby connection that is operated when an abnormality occurs in the connection between the IO node 913 and the IO device 917 indicated by a solid line.

  In such a computer system, when transferring data to an IO device, the calculation node transfers the data to an IO node connected to the target IO device by dimensional order routing. For example, data transferred by dimension order routing is first transferred along the X axis from the source node to a node whose X axis coordinates coincide with the X axis coordinates of the transfer destination node. Next, the data is transferred along the Y axis to the transfer destination node by changing the data transfer direction on the Y axis.

  For example, when the calculation node 901 in FIG. 21 transfers data to the IO device 918, the calculation node 901 transfers the data to the IO node 915 connected to the IO device 918 via the calculation nodes 902, 903, 907, and 911. To do. Note that the communication path selected by the dimension order routing is generally set in advance in each of the calculation nodes 901 to 912 and each of the IO nodes 913 to 916.

JP 2010-218364 A

  However, the above-described conventional technique has a problem that the availability of data transfer is not high.

  For example, when an abnormality occurs in a node or connection (interconnect) existing on a route selected by dimension order routing, the computation node cannot transfer data. In FIG. 21, a case where the connection between the calculation node 903 and the calculation node 907 is disconnected will be described as an example. The calculation node 901 tries to transfer data via the calculation node 903 according to the path selected by the dimension order routing, although it is connected to the IO node 915 through a path that bypasses the calculation node 903. As a result, the calculation node 901 cannot transfer data to the IO node 915.

  Further, in the conventional technique, when a failure occurs in the operational IO node connected to the IO device, the calculation node may not be able to transfer data. In FIG. 21, an example will be described in which when the computing node 906 transfers data to the IO device 918, an abnormality occurs in the IO node 915 that is the active system, and the IO node 914 is switched from the standby system to the active system.

  The calculation node 906 transfers the data to the calculation node 907, the calculation node 911, and the IO node 915 according to the route selected by the dimension order routing even when the active IO node is switched from the IO node 915 to the IO node 914. End up. Here, when the IO node 915 cannot transfer data to the IO node 914, the calculation node 906 cannot transfer data to the IO device 918.

  In addition, when the IO node 915 can transfer data to the IO node 914, the calculation node 906 can transfer data to the IO device 918, but through more nodes than the shortest path from the calculation node 906 to the IO node 914. Thus, data is transferred to the IO node 914. For example, when data is transferred from the computation node 906 to the IO node 914, the route having the smallest number of nodes is the route through the computation node 910. However, the computation node 906 follows a route that is set in advance. , Can't select this route. For this reason, when data can be transferred, there is a possibility that the data transfer by the calculation node is delayed.

  In one aspect, an object is to provide a management device, an information processing device, an information processing system, a data transfer method, a route selection program, and a transfer request program that can improve the reliability of data transfer.

  In the first proposal, the management device stores a communication state between information processing devices connected to each other or between the information processing device and the input / output device. Further, the management device stores a communication state between the input / output device that inputs / outputs data and the input / output processing device that inputs / outputs data to / from the input / output device in association with the identification information of the input / output processing device. Then, the management device receives a transfer request for requesting data transfer with the input / output device from any of the information processing devices. Further, the management device determines whether or not a communication path with the input / output device has been selected in response to the received transfer request. Here, if the management apparatus determines that the communication path with the input / output apparatus has been selected, the management apparatus specifies the transfer destination information processing apparatus or input / output processing apparatus from the selected communication path. On the other hand, when the management device determines that the communication path has not been selected, the management device, the communication state between the information processing devices connected to each other or between the information processing device and the input / output device, the input / output device, and the input / output device A communication path with the input / output device is selected based on the communication state between the input / output processing devices that input / output data to / from the device. Subsequently, the management apparatus identifies the information processing apparatus or input / output processing apparatus as the transfer destination based on the selected communication path. Then, the management device transmits an identifier for identifying the specified information processing device or input / output device to the information processing device that has issued the transfer request.

  The availability of data transfer can be improved.

FIG. 1 is a diagram illustrating a configuration example of a computer system. FIG. 2 is a block diagram showing the configuration of the computation node. FIG. 3 is a block diagram showing the configuration of the IO node. FIG. 4 is a block diagram showing the configuration of the management server. FIG. 5 is a diagram illustrating an example of the link management table. FIG. 6 is a diagram illustrating an example of the IO node management table. FIG. 7 is a diagram illustrating an example of the transfer data management table. FIG. 8 is a diagram illustrating an example of a communication path management table. FIG. 9 is a diagram illustrating an example of a computer system in which an abnormality occurs in the communication state between the computation node 11 and the IO node 15. FIG. 10 is a diagram illustrating an example of the link management table updated by the update unit. FIG. 11 is a diagram illustrating an example of a computer system in which an abnormality occurs in the IO node according to the first embodiment. FIG. 12 is a diagram illustrating an example of the IO node management table updated by the generation unit. FIG. 13A is a diagram illustrating an example of a communication path management table. FIG. 13B is a diagram illustrating an example of a communication path management table. FIG. 14 is a flowchart illustrating a processing procedure of data transfer processing to the IO device by the calculation node. FIG. 15 is a flowchart illustrating a processing procedure of processing by the management server. FIG. 16 is a flowchart illustrating the processing procedure of the update process. FIG. 17 is a flowchart illustrating a processing procedure of route selection processing. FIG. 18 is a block diagram illustrating the configuration of the management server according to the second embodiment. FIG. FIG. 19 is a diagram illustrating an example of a computer that executes a route selection program. FIG. 20 is a diagram illustrating an example of a computer that executes a transfer request program. FIG. 21 is a diagram illustrating a configuration example of a computer system adopting a cluster configuration.

  Hereinafter, embodiments of a management device, an information processing device, an information processing system, a data transfer method, a route selection program, and a transfer request program disclosed in the present application will be described in detail with reference to the drawings. Note that the present invention is not limited to the embodiments. Each embodiment can be appropriately combined within a range in which processing contents are not contradictory.

  Here, the configuration of the computer system according to the first embodiment, the configuration of the computation node, the configuration of the IO (Input Output) node, the configuration of the management server, the processing procedure of the processing by the computing node, the processing procedure of the processing by the management server, the embodiment The effects according to 1 will be described in order.

[Configuration of Computer System According to First Embodiment]
A computer system 1 illustrated in FIG. 1 includes calculation nodes 2 to 13, IO nodes 14 to 17, IO devices 20 to 22, and a management server 100. Here, the number of computing nodes, the number of IO nodes, and the number of IO devices included in the computer system 1 are merely examples, and are not limited thereto, and can be arbitrarily changed. The calculation nodes 2 to 13 and the IO nodes 14 to 17 are examples of information processing apparatuses. The management server 100 is an example of a management device.

  The management server 100 and each of the calculation nodes 2 to 13 are connected by, for example, a LAN (Local Area Network). Further, the management server 100 and each of the IO nodes 14 to 17 are connected by, for example, a LAN.

  Moreover, between each adjacent calculation node and each calculation node, between each adjacent calculation node and each IO node, and between each adjacent IO node and each IO node, it connects in mesh shape by an interconnect, for example. The In the example illustrated in FIG. 1, the calculation node 2 is connected to the calculation node 3 and the calculation node 6, respectively. The calculation node 11 is connected to the calculation node 7, the calculation node 10, the calculation node 12, and the IO node 15, respectively. The computer system 1 is described as being connected in a mesh shape, but is not limited to this, and may be connected in a torus shape, for example.

  The IO node 14 and the IO device 20 are connected by an IO bus such as Fiber Channel. The IO node 15 and the IO devices 20 and 21 are connected by a Fiber Channel. The IO node 16 and the IO device 22 are connected by a Fiber Channel or the like. The IO node 17 and the IO devices 21 and 22 are connected by a Fiber Channel or the like. Note that the IO node and the IO device are described as being connected by Fiber Channel, but the present invention is not limited to this, and may be connected by, for example, iSCSI.

  In the example of FIG. 1, the IO node has an active-passive configuration with respect to the IO device, and the connection between the active IO node and the IO device is indicated by a solid line, and the standby IO node and the IO device are connected to each other. Connections are indicated by dotted lines. Here, the active-passive configuration is a configuration in which an active node occupies the IO device during normal operation among a plurality of nodes sharing the IO device. A plurality of nodes sharing an IO device are divided into an active node and a standby node, and the active node and the standby node do not use the same IO device at the same time. During normal operation, only the active node uses the IO device, and when a failure occurs in the active node, the standby node uses the IO device. For example, for the IO device 20, the IO node 14 is an active node and the IO node 15 is a standby node. For the IO device 21, the IO node 15 is an active node, and the IO node 17 is a standby node. For the IO device 22, the IO node 17 is an active node, and for the IO device 22, the IO node 16 is a standby node. The IO device and the IO node in the computer system are not limited to the active-passive configuration. For example, the IO device and the IO node 15 may use the same IO device 20 at the same time in the active-active configuration. May be.

  Each of the calculation nodes 2 to 13 executes various calculations based on data input from the IO devices 20 to 22 and executes an application input from the IO devices 20 to 22.

  The calculation node monitors a communication state between the own device and the calculation node connected to the own device and a communication state between the own device and the IO node, and notifies the management server 100 of each monitored communication state. For example, the calculation node 12 monitors the communication state between the calculation node 8, the calculation node 11, and the calculation node 13 and the communication state with the IO node 16, and notifies the management server 100 of each monitored communication state. .

  In addition, when each of the calculation nodes 2 to 13 transfers data to and from the IO device, the calculation nodes 2 to 13 transmit a data transfer request indicating that the data transfer to and from the IO device is requested to the management server 100. For example, the case where the calculation node 7 shown in FIG. 1 transfers data to the IO device 21 will be described. In this case, the computation node 7 is not adjacent to the IO device 21. For this reason, when the calculation node 7 transfers data to the IO device 21, the data is relayed to other adjacent nodes. In the following description, a node that is a data transfer start source is appropriately described as a “transfer start source”, and an IO device that is a final data transfer destination is appropriately described as a “final transfer destination”.

  The calculation node 7 includes “the identifier of the device that is the transfer start source”, “the identifier that identifies the device itself”, “the identifier that identifies the data to be transferred”, and “the identifier of the IO device 21 that is the final transfer destination”. The transfer request is transmitted to the management server 100. It should be noted that the same value is stored in the “identifier of the device that is the transfer start source” and the “identifier that identifies the own device” transmitted by the calculation node 7 that is the transfer start source.

  Subsequently, the calculation node 7 receives from the management server 100 that it is the calculation node 11 as a transfer destination node to be a data relay destination node. Then, the computation node 7 transmits data to the computation node 11. Here, the data transmitted by the calculation node 7 includes “an identifier of a device that is a transfer start source”, “an identifier that identifies the device itself”, an “identifier that identifies data to be transferred”, and an “IO that is a final transfer destination”. The identifier of the device 21 ”is included.

  In addition, each of the calculation nodes 2 to 13 transmits a transfer request to the management server 100 when data is received from an adjacent calculation node. For example, a case where the calculation node 11 illustrated in FIG. 1 receives data to be transferred from the calculation node 7 that is the “transfer start source” to the IO device 21 that is the “final transfer destination” will be described.

  From the received data, the calculation node 11 “identifies the device that is the transfer start source”, “identifier that identifies the device itself”, “identifier that identifies the data to be transferred”, and “identifier of the IO device 21 that is the final transfer destination” Is extracted. Then, the calculation node 11 “the identifier of the calculation node 7 that is the transfer start source”, “the identifier that identifies the own device”, “the identifier that identifies the data to be transferred”, and “the identifier of the IO device 21 that is the final transfer destination” ”Is transmitted to the management server 100. The calculation node 11 receives from the management server 100 that it is the IO node 15 as a transfer destination node to be a data relay destination node. The computing node 11 then sends to the IO node 15 “the identifier of the computing node 7 that is the transfer start source”, “the identifier that identifies the own device”, “the identifier that identifies the data to be transferred”, and “the IO that is the final transfer destination”. The data including “the identifier of the device 21” is transmitted.

  Each IO node 14-17 controls input / output to each IO device 20-22. For example, the IO node 15 receives the data transferred from the calculation node 11 and outputs it to the IO device 21. Further, the IO node 15 outputs the data read from the IO device 21 to the calculation node 11.

  The IO node indicates a communication state between the own device and the IO node connected to the own device, a communication state between the own device and the calculation node connected to the own device, and a communication state between the own device and the IO device, respectively. The management server 100 is notified of the monitored communication state. For example, the IO node 15 monitors the communication state between the IO node 14 and the IO node 16, the communication state with the calculation node 11, and the communication state with the IO devices 20 and 21, respectively. Each state is notified to the management server 100.

  Each of the IO devices 20 to 22 is a storage device such as an HDD (Hard Disk Drive), and stores data and applications. Here, the IO devices 20 to 22 are described as storage devices such as HDDs, but are not limited to storage devices such as HDDs, and may be various sensors, printers, displays, and the like. In the following description, the calculation nodes 2 to 13 and the IO nodes 14 to 17 are simply described as the nodes 2 to 17 as appropriate.

  The management server 100 stores the notified communication states of the respective calculation nodes 2 to 13 and the IO nodes 14 to 17. In addition, the management server 100 stores communication states between a plurality of IO devices that store data and nodes that input and output data to the plurality of IO devices.

  In addition, the management server 100 executes the following processing when a data transfer request for requesting data transfer to any one of the plurality of IO devices 20 to 22 is received from any node. That is, the management server 100 selects a communication path between the IO devices based on the communication states between the plurality of nodes and the communication states between the nodes that input and output data to the IO devices and the IO devices. In addition, the management server 100 identifies a calculation node or an IO node as a transfer destination based on the selected communication path. Then, the management server 100 transmits an identifier for identifying the specified calculation node or IO node to the calculation node that issued the transfer request.

[Compute node configuration]
FIG. 2 is a block diagram illustrating the configuration of the calculation node according to the first embodiment. As shown in FIG. 2, the calculation node 2 includes a storage unit 2a and a control unit 2b. Since the configuration of the calculation nodes 2 to 13 is the same, only the configuration of the calculation node 2 will be described here, and the description of the configuration of the calculation nodes 3 to 13 will be omitted.

  The storage unit 2a is, for example, a storage device such as a semiconductor memory device or a hard disk, and stores various processing results executed by the own device and data and applications necessary for various processes executed by the own device. The storage unit 2a stores an identifier for identifying the device itself, an identifier for identifying an IO device that inputs and outputs data, and the like.

  The control unit 2b includes a first communication unit 2c, a second communication unit 2d, a link monitoring unit 2e, a transfer request unit 2f, and a transfer unit 2g. For example, the control unit 2b is an integrated circuit such as an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a central processing unit (CPU), or a micro processing unit (MPU). The control unit 2b has an internal memory for storing a control program, a program defining various processing procedures, and data necessary for various processing.

  The first communication unit 2c is connected to the management server 100 via a LAN, for example, and controls the exchange of information with the management server 100. For example, the first communication unit 2c transmits the communication state monitored by the link monitoring unit 2e to the management server 100. Further, the first communication unit 2 c receives the identifier of the transfer destination calculation node or IO node selected by the management server 100 from the management server 100. The first communication unit 2c outputs the received identifier of the transfer destination calculation node or IO node to the transfer request unit 2f.

  The second communication unit 2d is connected to an adjacent calculation node or an adjacent IO node by an interconnect, and controls the exchange of information with the adjacent calculation node or the adjacent IO node. For example, the second communication unit 2d transmits data to the transfer destination node based on the identifier of the transfer destination node received from the management server 100.

  The link monitoring unit 2e monitors the communication state with adjacent calculation nodes and adjacent IO nodes. The link monitoring unit 2e notifies the management server 100 of the monitored communication state via the first communication unit 2c. The link monitoring unit 2e is an example of a notification unit.

  When transferring data to / from an IO device that inputs and outputs data, the transfer request unit 2 f transmits a transfer request for requesting an identifier of an information processing device to which the data is to be transferred to the management server 100. For example, when transferring the data of its own device to an IO device that inputs and outputs data, a transfer request is generated, and the transfer request generated via the first communication unit 2 c is transmitted to the management server 100.

  For example, when the transfer request unit 2 f requests the IO device to transfer the data held by the own device, the “data ID”, “transfer start source ID”, “final transfer destination ID”, and “transfer source ID” A transfer request including “Data ID” included in the transfer request indicates an identifier of data to be transferred, and “Transfer start source ID” indicates an identifier of a calculation node that is a transfer start source of data identified by the data ID. “Final transfer destination ID” indicates the identifier of the IO device that is the final transfer destination of the data identified by the data ID, and “Transfer source ID” indicates the identifier of the own device. In the transfer request generated by the calculation node that is the data transfer start source, the same identifier is stored in “transfer start source ID” and “transfer source ID”.

  The transfer request unit 2f generates a transfer request when transferring data received from another device to an IO device that inputs and outputs data, and the transfer request generated via the first communication unit 2c is transmitted to the management server 100. Send to.

  For example, when the transfer request unit 2 f requests the IO device to transfer data received from another device, the “data ID”, “transfer start source ID”, “final transfer destination ID”, and “transfer source ID” A transfer request including “is generated. The “data ID”, “transfer start source ID”, and “final transfer destination ID” included in the transfer request include “data ID”, “transfer start source ID”, and “final transfer destination ID” included in the received data. Is stored in the same identifier. That is, the “data ID”, “transfer start source ID”, and “final transfer destination ID” included in the transfer request include “data ID” included in the transfer request generated by the calculation node that is the data transfer start source. , The same identifiers as “transfer start source ID” and “final transfer destination ID” are stored. The “transfer source ID” stores an identifier of the own device.

  In addition, the transfer request unit 2f, for the transfer request transmitted to the management server 100, the identifier of the transfer destination calculation node or the transfer destination IO node selected by the management server 100 based on the communication state notified by the link monitoring unit 2e. Is received from the management server 100 via the first communication unit 2c. Then, the transfer request unit 2f outputs the received transfer destination calculation node or IO node identifier to the transfer unit 2g. The transfer request unit 2f is an example of a transmission unit and a reception unit.

  The transfer unit 2g transfers the data to the calculation node or IO node having the identifier of the transfer destination calculation node or IO node received from the management server 100. For example, the transfer unit 2g sends the second communication unit 2d to the transfer destination calculation node or the transfer destination IO node based on the identifier of the transfer destination calculation node or the IO node received from the transfer request unit 2f via the first communication unit 2c. Send data via Here, the data transmitted by the transfer unit 2g includes “data ID”, “transfer start source ID”, “final transfer destination ID”, and “transfer source ID”. As described above, the transfer unit 2g includes the “data ID”, “transfer start source ID”, “final transfer destination ID”, and “transfer source ID” in the transfer destination calculation node or the transfer destination IO node. By transmitting the data, the node that received the data can determine the final transfer destination.

[IO node configuration]
FIG. 3 is a block diagram illustrating the configuration of the IO node according to the first embodiment. As shown in FIG. 3, the IO node 14 includes a storage unit 14a and a control unit 14b. Since the configuration of the IO nodes 14 to 17 is the same, only the configuration of the IO node 14 will be described here, and the description of the configuration of the IO nodes 15 to 17 will be omitted.

  The storage unit 14a is, for example, a storage device such as a semiconductor memory element or an HDD, and stores an identifier of the own device, an identifier of an IO device that inputs and outputs data, and the like.

  The control unit 14b includes a first communication unit 14c, a second communication unit 14d, a third communication unit 14e, a link monitoring unit 14f, a transfer request unit 14g, and a transfer unit 14h. For example, the control unit 14b is an integrated circuit such as an ASIC, FPGA, CPU, or MPU. The control unit 14b has an internal memory for storing a control program, a program that defines various processing procedures, and required data.

  The first communication unit 14c is connected to the management server 100 via, for example, a LAN, and controls the exchange of information with the management server 100. For example, the first communication unit 14c transmits the communication state monitored by the link monitoring unit 14f to the management server 100. Further, the first communication unit 14 c receives the identifier of the transfer destination calculation node or IO node selected by the management server 100 from the management server 100. The first communication unit 14c outputs the received identifier of the transfer destination calculation node or IO node to the transfer request unit 14g.

  The second communication unit 14d is connected to an adjacent calculation node or an adjacent IO node by an interconnect such as Tofu, and controls the exchange of information with the adjacent calculation node or the adjacent IO node. For example, the second communication unit 14 d transmits data to the transfer destination node based on the identifier of the transfer destination node received from the management server 100.

  The third communication unit 14e is connected to the IO device by, for example, Fiber Channel, and controls data input / output to / from the IO device. For example, the third communication unit 14e outputs the data received from the transfer unit 14h to the IO device.

  The link monitoring unit 14f monitors a communication state with an adjacent calculation node or an adjacent IO node. The link monitoring unit 14f notifies the management server 100 of the monitored communication state via the first communication unit 14c. The link monitoring unit 14f is an example of a notification unit.

  The transfer request unit 14g generates a transfer request that requests the management server 100 for an identifier of a transfer destination node when transferring data to an IO device that inputs and outputs data. Then, the transfer request unit 14g transmits the generated transfer request to the management server 100 via the first communication unit 14c. Further, the transfer request unit 14g transmits the identifier of the transfer destination calculation node or transfer destination IO node selected by the management server 100 based on the communication state notified by the link monitoring unit 14f in response to the transmitted transfer request. 100 from the first communication unit 14c. The transfer request unit 14g is an example of a transmission unit and a reception unit. The transfer request generated by the transfer request unit 14g includes “data ID”, “transfer start source ID”, “final transfer destination ID” and the like, as in the transfer request generated by the transfer request unit 2f of the computation node. , “Transfer source ID”.

  The transfer unit 14h transmits data to the transfer destination node via the second communication unit 14d based on the identifier of the transfer destination node received via the first communication unit 14c. Further, the transfer unit 14h outputs data received from the calculation node or the IO node to the IO device.

[Management server configuration]
Next, the configuration of the management server will be described with reference to FIG. FIG. 4 is a block diagram illustrating the configuration of the management server according to the first embodiment. As illustrated in FIG. 4, the management server 100 includes a storage unit 110 and a control unit 120.

  The storage unit 110 is a storage device such as a semiconductor memory device or an HDD, and includes a link management table 111, an IO node management table 112, a transfer data management table 113, and a communication path management table 114.

  The link management table 111 stores communication states between a plurality of nodes. FIG. 5 is a diagram illustrating an example of the link management table. For example, as illustrated in FIG. 5, the link management table 111 stores information in which “link ID”, “node a”, “node b”, and “communication state” are associated with each other.

  Here, the “link ID” stored in the link management table 111 includes information between adjacent calculation nodes and calculation nodes, between adjacent calculation nodes and IO nodes, and between adjacent IO nodes and IO nodes. Stores the identifier of the link to be connected. For example, “xxx1” and “xxx2” are stored in the “link ID”.

  “Node a” stores the identifier of one of the calculation nodes and IO nodes forming the link identified by the link ID. For example, “compute node 2”, “compute node 3”, and the like are stored in “node a”. Further, “node b” stores the identifier of the other of the calculation nodes and IO nodes forming the link identified by the link ID. For example, “node b” stores “calculation node 3”, “calculation node 4”, and the like.

  The “communication state” stores information indicating the communication state of the link identified by the link identifier. For example, “Normal” indicating that the communication state is normal and “Fall” indicating that the communication state is abnormal are stored in the “communication state”. Of the information stored as the link management table 111, information stored as “link ID, node a, node b” is set in advance by the administrator. The link management table 111 is an example of a first storage unit.

  In the example illustrated in FIG. 5, the link management table 111 indicates that the link identified by the link ID “xxx1” connects the calculation node 2 and the calculation node 3 and the link state is normal. Similarly, the link management table 111 indicates that the link identified by the link ID “xxx2” connects the calculation node 3 and the calculation node 4 and the link state is normal.

  The IO node management table 112 stores communication states between a plurality of IO devices that store data and nodes that input and output data to the plurality of IO devices. FIG. 6 is a diagram illustrating an example of the IO node management table. For example, as shown in FIG. 6, the IO node management table 112 stores information in which “link ID”, “IO node”, “IO device”, “communication state”, and “operation state” are associated with each other.

  Here, the “link ID” stored in the IO node management table 112 stores the identifier of the link connecting the IO node and the IO device. For example, “001”, “002”, and the like are stored in the “link ID”.

  The “IO node” stores the identifier of the IO node that forms the link identified by the link ID. For example, “IO node 14”, “IO node 15”, and the like are stored in “IO node”.

  The “IO device” stores an identifier of the IO device connected by the link identified by the link ID. For example, “IO device 20”, “IO device 21”, and the like are stored in the “IO device”.

  The “communication state” stores information indicating the communication state of the link. For example, “Normal” indicating that the communication state is normal and “Fall” indicating that the communication state is abnormal are stored in the “communication state”.

  Information indicating whether the link state is the standby system or the active system is stored in the “operation status”. For example, “Active” indicating that the operation state of the node is the active system and “Standby” indicating that the operation state of the node is the standby system are stored in the “operation state”.

  Specifically, in the IO node management table 112, the link identified by the link ID “001” connects the IO node 14 and the IO device 20, the link state is normal, and the IO node 14 is the active system. Indicates that Similarly, in the IO node management table 112, the link identified by the link ID “002” connects the IO node 15 and the IO device 20, the link state is normal, and the IO node 15 is the standby system. It shows that. Of the information stored as the IO node management table 112, information stored as “link ID”, “IO node”, “IO device”, and “operation state” is set in advance by the administrator. The IO node management table 112 is an example of a second storage unit.

  Returning to FIG. 4, the transfer data management table 113 stores an identifier of data requested to be transferred, a data transfer start source node, and a data final transfer destination node in association with each other. FIG. 7 is a diagram illustrating an example of the transfer data management table. For example, as illustrated in FIG. 7, the transfer data management table 113 stores information in which “data ID”, “transfer start source ID”, and “final transfer destination ID” are associated with each other.

  Here, the “data ID” stored in the transfer data management table 113 stores the identifier of the data requested to be transferred. For example, “xxx1”, “xxx2”, and the like are stored in the “data ID”. The data ID is given by the calculation node that is the data transfer start source.

  The “transfer start source ID” stores the identifier of the calculation node that is the transfer start source of the data identified by the data ID. For example, “calculation node 2”, “calculation node 7”, and the like are stored in “transfer start source ID”. The transfer start source ID is given by the calculation node that becomes the data transfer start source.

  The “final transfer destination ID” stores the identifier of the IO device that is the final transfer destination of the data identified by the data ID. For example, “IO device 20”, “IO device 21”, and the like are stored in the “final transfer destination ID”. The final transfer destination ID is given by the calculation node that is the data transfer start source.

  In the example illustrated in FIG. 7, the transfer data management table 113 indicates that the ID of data transferred to the IO device 20 with the calculation node 2 as the transfer start source is “xxx1”. Similarly, the transfer data management table 113 indicates that the ID of data transferred to the IO device 21 with the calculation node 7 as the transfer start source is “xxx2”.

  Returning to FIG. 4, the communication path management table 114 stores a communication path for transferring data requested to be transferred to a data transfer destination. FIG. 8 is a diagram illustrating an example of a communication path management table. For example, as shown in FIG. 8, the communication path management table 114 includes “data ID”, “communication path ID”, “transfer source ID”, “transfer destination ID”, “transfer start source ID”, “final transfer destination”. Information associated with each “ID” is stored. Note that the information stored in the “data ID”, “transfer start source ID”, and “final transfer destination ID” is the same as that in the transfer data management table 113, and thus the description thereof is omitted.

  Here, the “communication path ID” stores a transfer section identifier obtained by dividing a communication path for transferring data identified by the data ID to the IO device for each transfer section. For example, “xxx 2.01” or “xxx 2.02” for identifying the divided transfer sections is stored in the “communication path ID”.

  The “transfer source ID” stores an identifier of the transfer source node corresponding to the communication path ID. For example, “calculation node 7” and “calculation node 11” are stored in the “transfer source ID”. The “transfer destination ID” stores the identifier of the transfer destination node corresponding to the communication path ID. For example, “calculation node 11”, “IO node 15”, and the like are stored in “transfer destination ID”.

  In the example illustrated in FIG. 8, the communication path management table 114 indicates that the data transfer having the calculation node 7 as the transfer start source and the IO device 21 as the final transfer destination is identified by the data ID “xxx2”. In the communication path management table 114, there are three transfer sections identified as “xxx2.01”, “xxx2.02”, and “xxx2.03” as the communication path in the data transfer. Indicates to do. In the communication path management table 114, “xxx2.01”, which is the first transfer section, indicates that the calculation node 7 is a transfer source and the calculation node 11 is a transfer destination. The communication path management table 114 indicates that “xxx2.02” that is the second transfer section is a path from the calculation node 11 that is the transfer source to the IO node 15 that is the transfer destination. The communication path management table 114 indicates that “xxx2.03” that is the third transfer section is a path from the IO node 15 that is the transfer source to the IO device 21 that is the transfer destination.

  Returning to FIG. 4, the control unit 120 includes a reception unit 121, a transmission unit 122, an update unit 123, and a selection unit 124. For example, the control unit 120 is an integrated circuit such as an ASIC, FPGA, CPU, or MPU. The control unit 120 has an internal memory for storing a control program, a program defining various processing procedures, and data necessary for various processing.

  The receiving unit 121 receives information indicating a communication state of a link connecting each calculation node between the calculation node and the IO node adjacent to each calculation node, and a communication state of a link connecting the IO node and the IO device. Receives the information indicated. Then, the reception unit 121 outputs information indicating the communication state of the received link to the update unit 123.

  The receiving unit 121 receives information indicating that a communication state with an adjacent calculation node or IO node or a communication state with an IO device has been changed from any of a plurality of calculation nodes or IO nodes. When the reception unit 121 receives information indicating that the communication state of the computation node or the communication state of the IO node has been updated, the reception unit 121 outputs the information to the update unit 123.

  Further, the reception unit 121 receives a transfer request for requesting data transfer from any of the plurality of IO devices from either the computation node or the IO node. For example, the receiving unit 121 receives a transfer request indicating that a data transfer is requested from the computation node to the IO device. Then, the reception unit 121 outputs a transfer request indicating that a data transfer is requested to the calculation node IO device to the selection unit 124.

  The transmission unit 122 receives the identifier of the data transfer destination node selected by the selection unit 124, and transmits the received identifier of the data transfer destination node to the calculation node requesting the data transfer to the IO device.

  The updating unit 123 receives from the receiving unit 121 the link connecting each node from the communication state of the link connected to each computation node and the communication state of the link connected to each IO node, and the communication state of this link. Is generated.

  In the example illustrated in FIG. 5, when the update unit 123 receives a notification indicating that the communication state with the calculation node 3 is normal from the calculation node 2 via the reception unit 121, the update unit 123 changes to (A) in FIG. 5. Store information as shown. That is, the updating unit 123 stores “Normal” indicating that the communication state is normal in the “communication state” of the link ID “xxx1”.

  In the example illustrated in FIG. 6, when the update unit 123 receives a notification indicating that the communication state with the IO device 20 is normal from the IO node 14 via the reception unit 121, the update unit 123 returns to FIG. Store information as shown. That is, the update unit 123 stores “Normal” indicating that the communication state is normal in the “communication state” of the link ID “001”.

  Further, when the update unit 123 receives a notification indicating that the communication state of the link formed by each calculation node or the communication state of the link formed by each IO node has been updated from the reception unit 121, the link management table 111 and the IO node management table 112 are updated. The update process of the link management table 111 by the update unit 123 will be described with reference to FIGS. 9 and 10. FIG. 9 is a diagram illustrating an example of a computer system in which an abnormality occurs in the communication state between the calculation node 11 and the IO node 15, and FIG. 10 is a diagram illustrating an example of a link management table updated by the updating unit. .

  In the case of the computer system shown in FIG. 9, the updating unit 123 receives a notification from the receiving unit 121 indicating that the communication state of the link formed between the calculation node 11 and the IO node 15 is abnormal. In such a case, the update unit 123 indicates that the communication state is abnormal in the “communication state” corresponding to “link ID = xxx7” in the link management table 111 as shown in FIG. “Fall” is stored.

  Further, the update unit 123 updates the IO node management table 112 and the link management table 111 when receiving a notification indicating that an abnormality has occurred in the IO node from the IO node. The update processing of the IO node management table 112 and the link management table 111 when an abnormality occurs in the IO node by the update unit 123 will be described with reference to FIGS. 11 and 12. FIG. 11 is a diagram illustrating an example of a computer system in which an abnormality occurs in an IO node according to the first embodiment. FIG. 12 is a diagram illustrating an example of an IO node management table updated by a generation unit.

  As shown in FIG. 11, when a failure occurs in the IO node 15, the update unit 123 executes the following process. That is, as illustrated in FIG. 12A, the update unit 123 sets “Fall” indicating that the communication state is abnormal in “Communication state” corresponding to “Link ID = 003” in the IO node management table 112. "," And "Fall" indicating that there is an abnormality is stored in the "operation status". Further, as illustrated in FIG. 10A, the update unit 123 sets “Fall” indicating that the communication state is abnormal in the “communication state” corresponding to “link ID = xxx7” in the link management table 111. Is stored.

  When receiving the data transfer request, the selection unit 124 selects a communication path through which data can be transferred to the IO device based on the communication state of the link stored in the link management table 111 and the IO node management table 112. Hereinafter, an example of processing by the selection unit 124 will be described in detail. The selection unit 124 is an example of a specifying unit.

  For example, when receiving a data transfer request, the selection unit 124 determines whether a communication path has been selected. When the selection unit 124 determines that the communication path has been selected, the selection unit 124 extracts the transfer destination node from the path management table 114.

  Specifically, the selection unit 124 includes the data ID included in the data transfer request, the transfer start source ID for identifying the transfer start source calculation node, the final transfer destination ID for identifying the final transfer destination IO device, and the data transfer source node. And a transfer source ID for identifying. Then, the selection unit 124 determines whether or not a data ID that matches the extracted data ID exists in the transfer data management table 113.

  If the selection unit 124 determines that a data ID that matches the extracted data ID exists in the transfer data management table 113, the selection unit 124 determines that a communication path has been selected. Then, the selection unit 124 searches the route management table 114 and extracts a transfer destination ID of a record in which the data ID, the transfer start source ID, the final transfer destination ID, and the transfer source ID match. Thereafter, the selection unit 124 outputs the extracted transfer destination ID to the transmission unit 122.

  On the other hand, if the selection unit 124 determines that the data ID that matches the extracted data ID does not exist in the transfer data management table 113, the selection unit 124 determines that the communication path has not been selected. An example in which the selection unit 124 determines that the communication path has not been selected is a case where a transfer request is received from the transfer start source node.

  If the selection unit 124 determines that the communication path has not been selected, the selection unit 124 refers to the link management table 111 and the IO node management table 112 and can transfer data to the requested IO device from the calculation node. Select the shortest communication path.

  Specifically, when the selection unit 124 determines that the data ID that matches the extracted data ID does not exist in the transfer data management table 113, the selection unit 124 determines that the communication path has not been selected. Then, the selection unit 124 stores the extracted data ID, the transfer start source ID, and the final transfer destination ID in the transfer data management table 113 in association with each other.

  The selection unit 124 reads the IO node management table 112 and determines whether there is an active IO node that inputs / outputs data to / from the data transfer destination IO device. If the selection unit 124 determines that there is an operational IO node that inputs and outputs data to the IO device that is the data transfer destination, the selection unit 124 selects the operational IO node. Then, the selection unit 124 reads the link management table 111 and selects one shortest path from the data transfer start source calculation node to the selected IO node.

  Further, the selection unit 124 determines whether or not there is an abnormality in the link on the selected route, and generates the selected communication route as the route management table 114 when it is determined that there is no abnormality. Then, the selection unit 124 searches the route management table 114 and extracts a transfer destination ID of a record in which the data ID, the transfer start source ID, the final transfer destination ID, and the transfer source ID match. The selection unit 124 outputs the extracted transfer destination ID to the transmission unit 122.

  If the selection unit 124 determines that the link on the selected route is abnormal, the selection unit 124 selects a detour route that bypasses the link determined to be abnormal, and determines whether the detour route has been selected. . If the selection unit 124 determines that the detour route has been selected, the selection unit 124 determines whether there is an abnormality in the link on the selected detour route. If the selection unit 124 determines that the detour route cannot be selected, the selection unit 124 notifies the administrator of the abnormality.

  On the other hand, if the selection unit 124 determines that there is no active IO node that inputs / outputs data to / from the data transfer destination IO device, does the selection unit 124 have a standby IO node connected to the data transfer destination IO device? Determine whether or not. When the selection unit 124 determines that there is a standby IO node connected to the IO device that is the data transfer destination, the selection unit 124 sets the IO node that inputs and outputs data to the IO device stored in the IO node management table 112 as the standby system. Update to the IO node. Then, the selection unit 124 selects an IO node to be connected to the IO device, and selects one shortest path to the selected IO node.

  As illustrated in FIG. 11, when a failure occurs in the IO node 15 that inputs and outputs data to the IO device 21, the selection unit 124 executes the following processing. In other words, the selection unit 124 stores “Active” indicating the active system in the “operating state” corresponding to the link ID indicating the link connecting the IO node 17 and the IO device 21 in the IO node management table 112. To do.

  Next, communication path selection processing by the selection unit 124 when the calculation node 7 requests data transfer to the IO device 21 will be described with reference to FIGS. 1, 9, and 11. The selection unit 124 selects the IO node 15 as an IO node connected to the IO device 21. Then, as illustrated in FIG. 1, the selection unit 124 selects the calculation node 11 and the calculation node 15 as transfer nodes that transfer data from the calculation node 7 to the IO device 21 when there is no abnormality in the communication path. In this case, the selection unit 124 selects the calculation node 11 as a data transfer destination, and outputs the selected data transfer destination to the transmission unit 122. In this case, the selection unit 124 generates the communication path management table 114 shown in FIG. 8 based on the identifier of the selected transfer node.

  As shown in FIG. 9, when the link connecting the calculation node 11 and the calculation node 15 cannot be used due to a failure or the like, the selection unit 124 executes the following processing. That is, the selection unit 124 selects a communication path that bypasses the link formed between the calculation node 11 and the calculation node 15.

  For example, the selection unit 124 selects the IO node 15 as an IO node connected to the IO device 21, and the computation node 11, the computation node 12, and the IO node as transfer nodes that transfer data from the computation node 7 to the IO device 21. 16. Select the IO node 15. The selection unit 124 generates the communication path management table 114 shown in FIG. 13A based on the identifier of the selected transfer node. FIG. 13A is a diagram illustrating an example of a communication path management table. In this case, the selection unit 124 selects the calculation node 11 as a data transfer destination, and outputs the selected data transfer destination to the transmission unit 122. Note that the communication path selected by the selection unit 124 is not limited to this. For example, the selection unit 124 may select the calculation node 11, the calculation node 10, the IO node 14, and the IO node 15 as a transfer node that transfers data from the calculation node 7 to the IO device 21.

  As illustrated in FIG. 11, when an abnormality occurs in the IO node 15 and the link connecting the IO node 15 and the IO device 21 cannot be performed, the selection unit 124 executes the following processing. That is, the selection unit 124 selects the IO node 17 as an IO node connected to the IO device 21. For example, the selection unit 124 selects the calculation node 11, the calculation node 12, the calculation node 13, and the IO node 17 as a transfer node that transfers data from the calculation node 7 to the IO device 21. The selection unit 124 generates the communication path management table 114 illustrated in FIG. 13B based on the identifier of the selected transfer node. FIG. 13B is a diagram illustrating an example of a communication path management table. In this case, the selection unit 124 selects the calculation node 11 as a data transfer destination, and outputs the selected data transfer destination to the transmission unit 122. Note that the communication path selected by the selection unit 124 is not limited to this. For example, the selection unit 124 may select the calculation node 11, the calculation node 12, the IO node 16, and the IO node 17 as a transfer node that transfers data from the calculation node 7 to the IO device 21.

[Processing procedure of data transfer processing by compute node]
Next, a processing procedure of data transfer processing to the IO device by the calculation node according to the first embodiment will be described with reference to FIG. FIG. 14 is a flowchart illustrating a processing procedure of data transfer processing to the IO device by the calculation node. Here, the description will be made using the calculation node 2, but the processing described here is not limited to the processing by the calculation node 2, and the same processing is executed in other calculation nodes and IO nodes.

  If the computing node 2 determines that there is data to be transferred to any IO device (step S101, Yes), the management server 100 determines its own device ID, data ID, transfer start source ID, and final transfer destination ID. (Step S102). For example, when the calculation node 2 generates data to be transferred to any IO device, or receives data to be transferred to any IO device from an adjacent calculation node, the data to be transferred to the IO device Is determined to exist. Note that when the calculation node 2 itself generates data to be transferred to any IO device, the calculation node 2 generates a data ID by itself.

  The calculation node 2 receives the identifier of the transfer destination node from the management server 100 (step S103), and transmits data to the transfer destination node based on the received identifier of the transfer destination node (step S104). For example, the calculation node 2 transfers data including a data ID, an ID of the own device, a transfer start source ID, a final transfer destination ID, and the like to the transfer destination node. The calculation node 2 thereafter ends the process.

[Processing Procedure of Management Server According to First Embodiment]
Next, the processing procedure of the processing of the management server 100 according to the first embodiment will be described with reference to FIGS. Here, the processing procedure of the data transfer process by the management server 100 will be described with reference to FIG. 15, the processing procedure of the link management table and the IO node management table update process by the management server 100 will be described with reference to FIG. The processing procedure of the route selection process by the management server 100 will be described using FIG.

(Data transfer processing procedure)
FIG. 15 is a flowchart of the data transfer process performed by the management server according to the first embodiment. For example, the management server 100 executes processing upon receiving a data transfer request from any of the computation nodes to the IO device.

  As illustrated in FIG. 15, the reception unit 121 of the management server 100 receives a data transfer request to the IO device from the calculation node (step S <b> 201). Then, the selection unit 124 extracts the data ID (step S202), and determines whether or not the route corresponding to the extracted data ID has been selected (step S203). Here, the selection unit 124 determines whether or not a data ID that matches the extracted data ID exists in the transfer data management table 113. If the selection unit 124 determines that a data ID that matches the extracted data ID exists in the transfer data management table 113, the selection unit 124 determines that the communication path has been selected. On the other hand, if the selection unit 124 determines that the data ID that matches the extracted data ID does not exist in the transfer data management table 113, the selection unit 124 determines that the communication path has not been selected.

  When it is determined that the route has not been selected (No at Step S203), the selection unit 124 executes processing for selecting a communication route (Step S204). When the selection unit 124 determines that the route has been selected (step S203, Yes), or when the communication route is selected by the process of step S204 executed by the selection unit 124, the transmission unit 122 is selected. The transfer destination node is determined from the route, and the transfer destination is notified to the calculation node (step S205). The selection unit 124 then ends the process.

(Link management table and IO node management table update processing procedure)
FIG. 16 is a flowchart showing the processing procedure of the link management table and IO node management table update processing. The management server 100 executes processing upon receiving a notification indicating that the IO node is abnormal from the IO node in which the abnormality has occurred.

  As illustrated in FIG. 16, when the management server 100 receives an update notification from the IO node (Yes in step S301), the management server 100 updates the “communication state” in the IO node management table 112 (step S302). For example, when notified from the IO node 15 that an abnormality has occurred, the management server 100 stores “Fall” indicating the abnormality in the “communication state” of the IO node management table 112. When the management server 100 is notified of normality from the IO node 15 recovered from the abnormality, “Normal” indicating normality is stored in the “communication status” of the IO node management table 112 and “operation status” is stored. “Standby” is stored.

  Further, the management server 100 updates the communication state with the IO node in which an abnormality has occurred in the link management table 111 (step S303). For example, when notified from the IO node 15 that an abnormality has occurred, the management server 100 stores “Fall” indicating an abnormality in the “communication state” of the link formed with the IO node 15 in the link management table 111. . Further, when the management server 100 is notified of the normality from the IO node 15 recovered from the abnormality, the “normal” indicating “normal” in the “communication state” of the link formed with the IO node 15 in the link management table 111. Is stored.

(Route selection process)
FIG. 17 is a flowchart illustrating a processing procedure of route selection processing. This process corresponds to the process of step S204 in FIG. 15 and is executed when the selection unit 124 determines that the route corresponding to the extracted data ID has not been selected.

  As illustrated in FIG. 17, the selection unit 124 of the management server 100 extracts the data transfer source calculation node and the data transfer destination IO device (step S401), and generates the transfer data management table 113 (step S402). .

  The selection unit 124 reads the IO node management table 112 (step S403), and determines whether or not there is an active IO node connected to the data transfer destination IO device (step S404). If the selection unit 124 determines that there is no active IO node connected to the data transfer destination IO device (No in step S404), the selection unit 124 executes the following processing. That is, the selection unit 124 determines whether or not there is a standby IO node connected to the data transfer destination IO device (step S405).

  When the selection unit 124 determines that there is a standby IO node connected to the data transfer destination IO device (Yes in step S405), the update unit 123 updates the IO node management table 112 and the link management table 111 ( Step S406).

  When the selection unit 124 determines that there is an active IO node connected to the data transfer destination IO device (step S404, Yes), and when the IO node management table 112 and the link management table 111 are updated in step S406, The following processing is executed. That is, the selection unit 124 selects an IO node (step S407), and then reads the link management table 111 (step S408).

  Next, the selection unit 124 selects one shortest path from the data transfer source calculation node to the selected IO node (step S409), and determines whether there is an abnormality in the link or node on the selected path. (Step S410). If the selection unit 124 determines that there is no abnormality in the link or node on the selected route (step S410, No), the communication route management table 114 is generated (step S411), and the process ends.

  On the other hand, when the selection unit 124 determines that there is an abnormality in the link or node on the selected route (Yes in step S410), the selection unit 124 selects one of the shortest routes from the abnormal link or the route that bypasses the node. (Step S412). Then, the selection unit 124 determines whether or not the shortest route has been selected (step S413). If the selection unit 124 determines that the shortest path has been selected (step S413, Yes), the selection unit 124 proceeds to step S410 and determines whether there is an abnormality in the link or node on the selected shortest path.

  On the other hand, if the selection unit 124 determines that the shortest path cannot be selected (No at Step S413), the selection unit 124 notifies the abnormality (Step S414) and ends the process. If the selection unit 124 determines that there is no standby IO node connected to the data transfer destination IO device (No in step S405), the selection unit 124 notifies the abnormality (step S414) and ends the process.

[Effect of Example 1]
As described above, in the first embodiment, the selection unit 124 of the management server 100 notifies the transfer start source node of the identifier of the transfer destination node specified from the generated communication path management table 114. For this reason, even if an abnormality occurs in some links or nodes of the computer system 1, the calculation node that is the data transfer source can transfer the data to the IO device that is the transfer destination. For example, in the computer system 1, when a trouble occurs in a link or node on the communication path, the availability improving data transferred to improve the availability of the system can be transferred to the IO device.

  Although the management server 100 has been described as specifying the transfer destination node from the generated communication path management table 114 and transmitting the identifier of the specified transfer destination node to the transfer start source node, the present invention is not limited to this. Absent. For example, the management server 100 may generate the communication path management table 114 and transmit the generated communication path management table 114 to all the computation nodes and IO nodes on the communication path.

  In the first embodiment, the case where the entire input / output data is transferred together has been described. By the way, when the data to be written can be additionally written in multiple times on the transfer destination IO node, it is not necessary to transfer the entire data all at once. Further, when there is a data chunk management mechanism that divides data into a plurality of data, communication may be performed by allocating a time slot for each communication of each data chunk. Therefore, the second embodiment includes a data chunk management mechanism that divides data, and a time slot management mechanism that divides the right to use each link into time slots for each fixed time or transfer unit, and manages them as resources. An embodiment of the management server will be described.

[Configuration of Management Server According to Second Embodiment]
FIG. 18 is a block diagram illustrating the configuration of the management server according to the second embodiment. As illustrated in FIG. 18, the management server 200 includes a storage unit 110 and a control unit 220. In addition, here, about the function part which plays the same role as each part shown in FIG. 4, the detailed description is abbreviate | omitted by attaching | subjecting the same code | symbol.

  The control unit 220 includes a reception unit 121, a transmission unit 122, an update unit 123, a selection unit 124, a time slot management unit 225, and a data chunk management unit 226.

  The time slot management unit 225 manages the right to use each link divided into time slots for each fixed time or for each transfer unit as a communication resource. For example, the time slot management unit 225 receives a request for the right to use each link in the path from each computation node to the IO device when the total available bandwidth of the link is small with respect to the data amount, and uses each link. Assign rights. Also, the time slot management unit 225 outputs information on the right to use each link to the transmission unit 122. As a result, the transmission unit 122 transmits the link usage right assignment to the computation node that has requested the link usage right.

  The data chunk management unit 226 manages the association of data chunks, which are transfer units divided into predetermined sizes, with data before division, and the transfer destination of each data chunk. For example, the data chunk management unit 226 can effectively use the communication band for communication performed by the own device and communication performed by another node by managing time-slotted time slots.

  As described above, the management server 200 divides the data and divides the right to use each link into a time slot for each fixed time or for each transfer unit. As a result, the computation node becomes congested when data transfer exceeding the communication band is started. It is possible to prevent a decrease in effective communication bandwidth due to the above. For example, the management server 200 is effective in transferring availability data in which a large amount of data to be transferred occurs in a short time.

  By the way, this invention may be implemented with a various different form other than the Example mentioned above. Thus, in the third embodiment, another embodiment included in the present invention will be described.

(System configuration etc.)
Of the processes described in the first and second embodiments, all or a part of the processes described as being automatically performed may be performed manually. Alternatively, all or part of the processing described as being performed manually can be automatically performed by a known method. In addition, the processing procedures, control procedures, and specific names shown in the text and drawings can be arbitrarily changed unless otherwise specified.

  Further, the information stored in the illustrated storage unit is only an example, and it is not always necessary to store the information as illustrated.

  The selection unit 124 determines whether or not a route corresponding to the extracted data ID has been selected by determining whether or not a data ID that matches the extracted data ID exists in the transfer data management table 113. However, the present invention is not limited to this. For example, the selection unit 124 extracts “transfer start source ID” and “transfer source ID” included in the transfer request, and determines whether “transfer start source ID” and “transfer source ID” match. Thus, it may be determined whether or not the route has been selected. If the selection unit 124 determines that the “transfer start source ID” matches the “transfer source ID”, the selection unit 124 determines that the route is not selected. On the other hand, if the selection unit 124 determines that the “transfer start source ID” does not match the “transfer source ID”, the selection unit 124 determines that the route has been selected. That is, the selection unit 124 determines that a route is not selected when a transfer request is received from a transfer start source, and has selected a route when a transfer request is received from a node that is not a transfer start source. Is determined.

  In addition, although the link between the adjacent calculation node and the calculation node, between the adjacent calculation node and the IO node, and between the adjacent IO node and the IO node has been described as a physical link, it is not limited to this. It is not something. For example, a link between an adjacent computation node and a computation node, between an adjacent computation node and an IO node, and between an adjacent IO node and an IO node may be a virtual channel.

  Further, the order of processing in each step of each processing described in each embodiment may be changed according to various loads and usage conditions. For example, step S302 and step S303 shown in FIG. 16 may be executed simultaneously.

  Each illustrated component is functionally conceptual and does not necessarily need to be physically configured as illustrated. For example, the update unit 123 and the selection unit 124 included in the management server 100 may be integrated. Furthermore, all or a part of each processing function performed in each device may be realized by a CPU and a program that is analyzed and executed by the CPU, or may be realized as hardware by wired logic.

(program)
By the way, the various processes described in the above embodiments can be realized by executing a program prepared in advance on a computer system such as a personal computer or a workstation. Therefore, in the following, an example of a computer that executes a program having the same function as the above embodiment will be described.

  FIG. 19 is a diagram illustrating an example of a computer that executes a route selection program. As illustrated in FIG. 19, the computer 300 includes an input device 310 that receives data and various settings from a user, and an output device 320 that notifies the status of the computer. The computer 300 includes a network interface 330 that transmits and receives data to and from other devices, a medium reading device 340, a RAM (Random Access Memory) 350, a ROM (Read Only Memory) 360, a CPU 370, and a bus 380. Each device 310 to 370 is connected to a bus 380.

  Here, as shown in FIG. 19, the ROM 360 stores in advance a route selection program 361 that exhibits the same functions as the reception unit 121, the transmission unit 122, the update unit 123, and the selection unit 124 shown in FIG. 4. ing. Then, the CPU 370 reads the route selection program 361 from the ROM 360 and executes it as the route selection process 371. That is, the route selection process 371 performs the same operations as those of the reception unit 121, the transmission unit 122, the update unit 123, and the selection unit 124 illustrated in FIG.

  FIG. 20 is a diagram illustrating an example of a computer that executes a transfer request program. As illustrated in FIG. 20, the computer 400 includes an input device 410 that receives data and various settings from a user, and an output device 420 that notifies the status of the computer. The computer 400 also includes a network interface 430 that transmits and receives data to and from other devices, a medium reading device 440, a RAM 450, a ROM 460, a CPU 470, and a bus 480. Each device 410 to 470 is connected to a bus 480.

  Here, as shown in FIG. 20, the ROM 460 stores in advance a transfer request program 461 that exhibits the same functions as the link monitoring unit 2e, transfer request unit 2f, and transfer unit 2g shown in FIG. . Then, the CPU 470 reads the transfer request program 461 from the ROM 460 and executes it as the transfer request process 471. That is, the transfer request process 471 executes the same operations as those of the link monitoring unit 2e, the transfer request unit 2f, and the transfer unit 2g shown in FIG.

  By the way, the above-described route selection program 361 and transfer request program 461 are not necessarily stored in the ROM. For example, it may be stored in a “portable physical medium” such as a flexible disk (FD), a CD-ROM, an MO disk, a DVD disk, a magneto-optical disk, or an IC card inserted into the computers 300 and 400. Good. Further, it may be stored in a “fixed physical medium” such as a hard disk drive (HDD) provided inside or outside the computer 300 or 400. Further, it may be stored in “another computer system” connected to the computers 300 and 400 via a public line, the Internet, a LAN (Local Area Network), a WAN (Wide Area Network), or the like. Then, the computers 300 and 400 may read out and execute the program from these.

  That is, the program is stored in a recording medium such as the above-mentioned “portable physical medium”, “fixed physical medium”, “communication medium”, and the like so as to be readable by a computer. The computers 300 and 400 implement the same functions as in the above-described embodiments by reading and executing the program from such a recording medium. Note that the program referred to in the other embodiments is not limited to being executed by the computers 300 and 400. For example, the present invention can be similarly applied to a case where another computer system or server executes a program or a case where these programs cooperate to execute a program.

DESCRIPTION OF SYMBOLS 1 Computer system 2-13 Computation node 2a Storage part 2b Control part 2c 1st communication part 2d 2nd communication part 2e Link monitoring part 2f Transfer request part 2g Transfer part 14-17 IO node 14a Storage part 14b Control part 14c 1st communication Unit 14d second communication unit 14e third communication unit 14f link monitoring unit 14g transfer request unit 14h transfer unit 20-22 IO device 100 management server 110 storage unit 111 link management table 112 IO node management table 120 control unit 121 reception unit 122 transmission Part 123 update part 124 selection part

Claims (5)

First storing a communication state indicating whether communication between the information processing apparatuses connected to each other or between the input / output processing apparatus that inputs / outputs data to / from the input / output apparatus that inputs / outputs data is normal or abnormal A storage unit;
A second storage unit that stores a communication state indicating whether communication between the input / output device and the input / output processing device is normal or abnormal, in association with an identifier of the input / output processing device;
A receiving unit that receives a transfer request for requesting data transfer with the input / output device from any of the information processing devices;
In response to the transfer request received by the receiver, it is determined whether a communication path with the input / output device has been selected, and it is determined that a communication path with the input / output device has been selected. If the information processing device or input / output processing device connected to the information processing device that issued the transfer request is specified, the information processing device or input / output processing device that is the data output destination in the selected communication path is specified, When it is determined that the communication path has not been selected, the communication path between the input / output device is selected based on the communication state stored in the first storage unit and the communication state stored in the second storage unit Then, among the information processing devices or input / output processing devices connected to the information processing device that issued the transfer request, the information processing device or input / output processing device that is the data output destination in the selected communication path is specified. And tough,
A management device comprising: a transmission unit that transmits an identifier for identifying an information processing device or an input / output processing device specified by the specifying unit to the information processing device that has issued the transfer request. In the management device,
When the abnormality occurs in the input / output processing device that inputs / outputs data to / from the transfer request destination input / output device or the connection between the input / output device and the input / output processing device, the specifying unit It is determined whether there is another input / output processing device that inputs / outputs data to / from the input / output device, and when it is determined that there is the other input / output processing device, the second storage is associated with the input / output device. Update the identifier of the input / output processing device that inputs and outputs data to the input / output device stored in the unit to the identifier of the other input / output processing device, and select a communication path between the input / output device, 2. The management apparatus according to claim 1, wherein an information processing apparatus or an input / output processing apparatus as a data transfer destination is specified based on the selected communication path. The management device further includes:
The reception unit is normal or abnormal in communication with any other information processing device or input / output processing device adjacent to the information processing device, or from any input / output processing device. When the information indicating that the normality or abnormality of the communication between the information processing device or the input / output device adjacent to the input / output processing device is changed, the communication state stored in the first storage unit and The management apparatus according to claim 1, further comprising an update unit that updates a communication state stored in the second storage unit. A plurality of information processing devices connected in a mesh shape or a torus shape, an input / output device for inputting / outputting data from the plurality of information processing devices, and an input / output processing device for inputting / outputting data to / from the input / output device An information processing system having a management device connected to each of the plurality of information processing devices,
The management device
A first storage unit that stores a communication state indicating whether communication between information processing apparatuses connected to each other or between the input / output processing apparatus and the information processing apparatus is normal or abnormal;
A second storage unit that stores a communication state indicating whether communication between the input / output device and the input / output processing device is normal or abnormal, in association with an identifier of the input / output processing device;
A receiving unit that receives a transfer request for requesting data transfer with the input / output device from any of the information processing devices;
In response to the transfer request received by the receiver, it is determined whether a communication path with the input / output device has been selected, and it is determined that a communication path with the input / output device has been selected. If the information processing device or input / output processing device connected to the information processing device that issued the transfer request is specified, the information processing device or input / output processing device that is the data output destination in the selected communication path is specified, When it is determined that the communication path has not been selected, the communication path between the input / output device is selected based on the communication state stored in the first storage unit and the communication state stored in the second storage unit Then, among the information processing devices or input / output processing devices connected to the information processing device that issued the transfer request, the information processing device or input / output processing device that is the data output destination in the selected communication path is specified. And tough,
An identifier for identifying the information processing device or input / output processing device specified by the specifying unit, and a transmission unit that transmits the identifier to the information processing device that issued the transfer request
The information processing apparatus includes:
A notification unit that monitors a communication state with another connected information processing apparatus and notifies the management apparatus of the monitored communication state;
When transferring data to / from an input / output device that inputs / outputs data, a transmission unit that transmits to the management device a transfer request that requests an identifier of the information processing device or input / output processing device to which the data is to be transferred ,
A receiving unit that receives, from the management device, an identifier of an information processing device or an input / output processing device that is a transfer destination selected by the management device based on the notified communication state in response to the transfer request transmitted by the transmission unit;
An information processing system comprising: an information processing device having an identifier of a transfer destination information processing device received from the management device, or a transfer unit that transfers the data to an input / output processing device. A plurality of information processing devices connected in a mesh shape or a torus shape, an input / output device for inputting / outputting data from the plurality of information processing devices, and an input / output processing device for inputting / outputting data to / from the input / output device A data transfer method in an information processing system having a management device connected to each of the plurality of information processing devices,
The information processing apparatus is
When transferring data to and from the input / output device, send a transfer request to the management device requesting an identifier of the information processing device or input / output processing device to which the data is to be transferred,
The management device is
A transfer request for requesting data transfer with the input / output device is received from any of the information processing devices,
In response to the received transfer request, it is determined whether a communication path with the input / output device has been selected, and when it is determined that a communication path with the input / output device has been selected, Among the information processing devices or input / output processing devices connected to the information processing device that has issued the transfer request, the information processing device or input / output processing device that is the data output destination in the selected communication route is specified, and the communication route The communication status indicating whether communication between information processing apparatuses connected to each other or between the input / output processing apparatus and the information processing apparatus is normal or abnormal, and the input / output apparatus and the previous An information processing device connected to the information processing device that has transmitted the transfer request by selecting a communication path with the input / output device based on the communication state indicating whether the communication between the entry output processing devices is normal or abnormal, or In the output processing unit identifies an information processing apparatus or output processing apparatus as the output destination of the data in the selected communication path,
An identifier for identifying the specified information processing device or input / output device is transmitted to the information processing device that issued the transfer request,
The information processing apparatus is
In response to the transmitted transfer request, the identifier of the transfer destination information processing device or input / output processing device selected by the management device based on the notified communication state is received from the management device,
A data transfer method comprising: transferring the data to an information processing apparatus or an input / output processing apparatus having an identifier of a transfer destination information processing apparatus received from the management apparatus.

Images